Re: negative object sizes?

On fre, 2007-11-23 at 09:55 +0900, Adrian Chadd wrote:
> I've been fiddling with Squid-2.HEAD and google maps, and I've noticed that
> weird stuff happens.
>
> Here's the log entries:
>
> 1195739570.596 SWAPOUT 00 00034F1B BEAF584D4381EE1DCAF38A162BAA0224 200 1195739570 1103259488 1227275570 image/jpeg 11389/11389 GET http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq
> 1195777355.041 RELEASE 00 00034F1B 748D838AAD5499315AEE444C306BE06C 200 1195739570 1103259488 1227275570 image/jpeg 11389/-243 GET http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq
> 1195777356.133 RELEASE -1 FFFFFFFF C208FB77DE2B773EFFBCF7DA928D1DF7 200 1195777354 1103259488 1227313354 image/jpeg 11389/11389 GET http://kh2.google.com.au/kh?n=404&v=23&t=ttrtq
>
> Now, the log entries in access.log so TCP_MISS, then TCP_REFRESH_MISS.
>
> I'll sort out the revalidation issue, but I note that there's a negative object
> size there - that -243 is in store_log.c:
>
> mem->inmem_hi - mem->reply->hdr_sz,
>
> Any insights into that?

I would guess the object was purged from memory before RELEASE, and that
our changes in the store client wrt how headers on swapins is handled
makes a memobject to be attached to the StoreEntry again but without any
data..

Regards
Henrik